Atmospheric Hazard Modelers Face Predictability Issues

Posted: July 7, 2008 at 1:00 am, Last Updated: November 30, -0001 at 12:00 am


This model shows the path of smoke from California wildfires. (Roll over image to display video control.)
Courtesy of Zafer Boybeyi

By Tara Laskowski

When atmospheric scientists, researchers and government officials from around the world gather this week to attend the 12th annual George Mason University Conference on Transport and Dispersion Modeling, they will discuss issues that Zafer Boybeyi, an associate professor in the Department of Earth Systems and Geoinformation Sciences, and his team deal with daily.

Zafer Boybeyi
Zafer Boybeyi
Creative Services photo

Boybeyi works to improve the modeling of hazardous atmospheric releases such as intentional and/or unintentional chemical, biological, radiological and nuclear releases into the atmosphere, as well as dust storms, volcanic eruptions, wildfires and biomass burning. He also is concerned with modeling severe weather such as hurricanes, tornadoes, thunderstorms, blizzards, heavy precipitation events and floods. If hazards occur, Boybeyi wants to track and model them.

Boybeyi has hosted and organized the conference for many years. This week, he will join the more than 150 attendees to look at such issues as the accuracy and predictability of numerical models on different scales. They will also discuss hazard assessment and emergency response tool developments and applications, model developments, real-time predictions and hazard physics.

Tropical Storm (Hurricane) Modeling

Boybeyi and his team currently are focusing their attention on hurricanes. With another hurricane season just beginning, it is important not only to predict what this year will look like, but also to look at the past to determine why certain seasons were rougher than others.


A model of Hurricane Katrina. (Roll over image to display video control.)
Courtesy of Zafer Boybeyi

With a grant from the National Science Foundation, Boybeyi is working to improve hurricane modeling and determine what conditions cause hurricanes to form more frequently. To do this, he and his team must take into account all kinds of factors: sea surface temperature, wind shear conditions, atmospheric moisture and even dust storm activities in Africa.

And then there’s what happened in 2005.

With several major destructive storms such as Hurricanes Katrina, Rita and Wilma, 2005 was an anomaly by way of hurricane activity. Boybeyi and his team looked at 150 years of hurricane seasons and found that 2005 was by far the worst.

“Why in 2005 was the hurricane activity so bad?” asks Boybeyi. “Many predicted that it was due to global warming, that things could continue to get worse. But then in 2006 and 2007, there were no significant storm activities. So we just need to understand this better.”

Super Models

A scientist is only as good as his model, you might say.

The challenge of predicting atmospheric hazards with as much precision as possible is one that Boybeyi and other scientists continue to grapple with, and it only gets more difficult as the scale of interest decreases. Imagine modeling a gas leak inside a building, taking into account the ventilation system, windows, doorways and temperature. Now imagine trying to model a biohazard spill on a campus — or better yet, the climate changes of the entire Earth — and you can begin to see the monumental task these scientists face.

In addition, the randomness of movement and the multiscale nature of the atmosphere are highly variable, making it impossible to ever be 100 percent accurate because model simulation tries to replicate the real world based on a set of assumptions and conceived representations of reality. Errors and uncertainties associated with model simulations always exist. Another, more pressing issue is that computer technology has not caught up to scientific ideas.

“There are simply not fast enough computers in existence yet to create more precise operational models on a large scale,” says Boybeyi.

Nevertheless, numerical modeling has seen considerable progress in the past several decades, Boybeyi says. Some of this is due to ever-increasing computational power and the use of satellite-derived remotely sensed data. Despite these developments, the models used in hazard prediction still remain scale-specific. There is general consensus among the modelers that the simulation of the atmosphere at all relevant scales together is an intractable problem.

Global to Local Scale Atmospheric Hazard Modeling

One of the most innovative modeling techniques, the Solution Adaptive Modeling Technique (SAMT), is being used at Mason. The technique was developed at Science Applications International Corporation by a research team that included Boybeyi when he worked for the company several years ago.

The SAMT allows scientists to study a targeted local area. For example, if one has bad eyesight and is sitting in front of a two-page newspaper spread, a magnifying glass would help target a specific article and make it easier to read. In the same way, if mapping the entire East Coast of the United States, the SAMT allows researchers to target the areas where a hurricane will most likely hit and show them in high-resolution for better accuracy.

“Our true understanding comes from the simulation of interaction between the various scales of motion in the atmosphere,” says Boybeyi. “Tools like this will help us to better understand the scale interaction.

“You can have a model and show pretty pictures, but it’s the science behind it that matters.”

Write to at